Add dataset card description and task category

README.md

@@ -50,4 +50,13 @@ configs:
   data_files:
   - split: train
     path: data/train-*
+task_categories:
+- question-answering
 ---
+
+# FrameBench
+
+FrameBench is a benchmark grounded in frame semantics. It consists of multiple-choice questions that test whether language models can distinguish the frames evoked by the same verb across contexts. This repository contains the English portion of FrameBench; the Japanese portion is available at [cl-nagoya/jFrameBench](https://huggingface.co/datasets/cl-nagoya/jFrameBench).
+
+- **Paper:** [FrameBench: A Language Understanding Benchmark Based on Frame Semantics](https://huggingface.co/papers/2609.03370)
+- **Code:** [https://github.com/SasanoLab/FrameBench](https://github.com/SasanoLab/FrameBench)